Marine radioisotope gamma-ray spectrum analysis method based on Geant4 simulation and MLP neural network

AbstractThe ability to precisely predict the volume fraction percentage of the different phases flowing in a pipe plays an important role in the oil, petroleum and other industries. In this research, the volume fraction percentage was measured precisely in water-gasoil-air three-phase flows by using a single pencil beam gamma ray attenuation technique and multilayer perceptron (MLP) neural network. The volume fraction percentage determination in three-phase flows requires least two gamma radioactive sources with different energies while in this study, we used just a 137Cs source (with the single energy of 662 keV) and a NaI detector. Also, in this work, the MLP neural network in MATLAB software was implemented to predict the volume fraction percentage. The experimental setup provides the required data for training and testing the network. Using this proposed method, the volume fraction was predicted in water-gasoil-air three-phase flows with mean relative error percentage less than 6.95 %. Also, the root mean square error was calculated 2.60. The set-up used is simpler than other proposed methods and cost, radiation safety and shielding requirements are minimized.

Download Full-text

A Spectrum Analysis Method Based on Neural Network

Applied Mechanics and Materials ◽

10.4028/www.scientific.net/amm.543-547.2846 ◽

2014 ◽

Vol 543-547 ◽

pp. 2846-2849

Author(s):

Wei Ping Cui ◽

Zhi Wen Cao

Keyword(s):

Neural Network ◽

Spectrum Analysis ◽

Amplitude Spectrum ◽

Complex Multiplication ◽

Signal Amplitude ◽

Least Square ◽

Analysis Method ◽

Recursive Least Square ◽

Software And Hardware ◽

Filtering Function

This paper presents a spectrum analysis method using recursive least square algorithm to train the weights of Fourier Basis Functions (FBF) neural network, according to the weight to obtain the signal amplitude spectrum and phase spectrum. The method does not involve complex multiplication and addition operations, convenient for software and hardware, especially suitable for DSP software and hardware implementation. The simulation results show that, this method is not only high precision, fast calculation speed, but also has the noise filtering function, is a kind of effective method for spectrum analysis.

Download Full-text

A Resolution Enhancing Algorithm for Gamma-Ray Spectrum Based on Blind Deconvolution and Lp-Norm Sparsity Constraint

Volume 3: Nuclear Fuel and Material, Reactor Physics and Transport Theory; Innovative Nuclear Power Plant Design and New Technology Application ◽

10.1115/icone25-66923 ◽

2017 ◽

Author(s):

Xinpeng Li ◽

Sheng Fang ◽

Hong Li

Keyword(s):

Spectrum Analysis ◽

Gamma Ray ◽

Blind Deconvolution ◽

System Response ◽

High Resolution Spectrum ◽

Reconstruction Method ◽

Analysis Method ◽

Response Matrix ◽

Lp Norm ◽

Sparsity Constraint

Current gamma-ray spectrum analysis method uses a preset system response matrix to improve the resolution of gamma-ray spectrum. However, the system response matrix may not be available or biased due to limitation of experiment conditions, which can degrade the accuracy of gamma-ray spectrum analysis. To solve the problem, a new reconstruction method based on blind deconvolution and sparsity constraint is proposed to improve the resolution of gamma-ray spectrum in this study. The proposed method models the modulation of spectrometer as a convolution operation and reconstructs the high resolution spectrum as well as the convolution kernel simultaneously. Lp-norm based sparsity constraint is imposed to stabilize the demodulation of spectrometer and reduce the background oscillations, so that the resolution can be enhanced. The results of both numerical simulation and experiments demonstrate that the proposed method can effectively improve the resolution of gamma-ray spectrum and reduce background oscillations without any aid of system response matrix.

Download Full-text